CANDIE: A New Version of the DieCAST Ocean Circulation Model

Abstract

The development and verification of a new version of the DieCAST ocean circulation model to be referred to as CANDIE (Canadian Diecast) are considered. Both CANDIE and DieCAST have many features in common with the well-known Modular Ocean Model (MOM) of the Geophysical Fluid Dynamics Laboratory. Of particular relevance to the present study are the rigid-lid approximation and the use of standard Cartesian coordinates. The DieCAST formulation in terms of the surface pressure, rather than the volume transport streamfunction, is also used in CANDIE to reduce numerical sensitivity to ocean depth variations. The major difference between MOM and DieCAST is the use of a mixed C and A grid formulation in DieCAST rather than the B grid formulation used in MOM. CANDIE differs from DieCAST in the use of a standard C grid formulation and a reduction in the magnitude of the time truncation error associated with the implicit treatment of the Coriolis force. The implementation of the rigid-lid approximation is retained. To assess the reliability of both DieCAST and CANDIE, the authors have applied these models to a problem used by Haidvogel and Beckmann in a comparison of several different model formulations. The tests include the influence of a steep-sided coastal canyon that represents a significant challenge for the step topography of Cartesian coordinate models. Haidvogel and Beckmann?s tests show general agreement between models based on topography-following coordinates, but significantly different results were obtained with MOM. The results of DieCAST for the homogeneous test case also differ substantially from those of the σ-coordinate models, largely due to dissipation associated with low-order interpolations used adjacent to solid boundaries in DieCAST. However, the results of CANDIE are in good agreement with those of the σ-coordinate models for both homogeneous and stratified coastal canyon experiments. These results clearly demonstrate that the differences found for both MOM and DieCAST are not due to intrinsic limitations associated with the use of Cartesian coordinates.